1

I have a source of data that follows this bivariant Gaussian distribution:

enter image description here

The same source of data produces this 2-dimensional histogram:

enter image description here

Computing the 2d histogram is much less expensive however the source is better described using a Gaussian. My question is: is there a way of approximating the 2d histogram to the bivariate Gaussian shown above? Maybe some sort of filtering using the mean and covariance from the source?

Both distribution maps (i.e. the images from above) are generated from values ranging [0,255]. In other words, my source outputs integer values between 0 and 255.

karl71
  • 71
  • 6
  • 1
    Do note that bin choice greatly affects the "shape" of the histogram and thus it can fit to different distributions. –  Jan 05 '17 at 11:05
  • @gunbl4d3, reducing the number of bins results in a lower resolution representation of the data. Not sure about the "greatly affects"... – karl71 Jan 05 '17 at 11:22
  • What do you mean by "approximating the 2d histogram to the Gaussian"? – Tim Jan 05 '17 at 17:50
  • 1
    @gunbl4d3 is not necessarily talking about resolution, but also things like [this](http://stats.stackexchange.com/a/51753/127790) (which I would call [aliasing](https://en.wikipedia.org/wiki/Aliasing)). – GeoMatt22 Jan 05 '17 at 18:00
  • @Tim, I'd like to find a way of making the 2d histogram look like the 2d Gaussian. (see the images in the question body). – karl71 Jan 06 '17 at 11:06

0 Answers0